3. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Published in- world journal of neurosurgery
Published year- 2020
Impact factor- 1.723
4. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Introduction
Traumatic brain injury (TBI) is an important medical, socioeconomic, and universal
health problem that occurs at all ages and in every population.
It is a significant cause of morbidity and mortality among young people, and the
incidence is increasing among people who are 65 years of age
The World Health Organization estimates that by 2020, traffic accidents will be the
third most frequent cause of death and invalidity
In high-income countries, the incidence of TBI is increasing because of falls, which
also causes a shift in the mean age of patients with neurotrauma toward elderly
5. Department of Neurosurgery
Tribhuvan University Teaching Hospital
The outcome after TBI is variable, but often patients cannot return to their previous
position in society
The preferred method for evaluating the outcome after TBI is the Glasgow Outcome Scale
(GOS), with scores ranging from 5 (i.e., low disability) to 1 (i.e., death)
Statistical models use multiple variables to estimate outcome and are based on the average
results of patients
The CRASH calculator was developed based on the Medical Research Council CRASH
trial, which investigated the effect of corticosteroids on death and disability of patients
with TBI
6. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Prognostic variables included
age,
sex,
cause of trauma,
time between trauma and randomization,
Glasgow Coma Scale (GCS) score,
pupil reactivity,
computed tomography (CT) scan findings,
presence of significant extracranial injuries, and
income level of the country where trauma occurred.
7. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Web-based and smartphone versions of the calculator were developed to make it
internationally accessible for all physicians
This study aimed to assess the accuracy of the calculator and to validate it by
applying it to patients with neurotrauma from 2010 to 2014 at Ghent University
Hospital, Belgium.
8. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Materials and methods
Data collection
This cross-sectional observational study was performed at Ghent University
Hospital, Ghent, Belgium.
The study received approval of the Ethical Committee of the Ghent University
Hospital
Both the data to calculate the CRASH score and the outcome data were
collected retrospectively from electronic patient records.
The CRASH scores were calculated prospectively for 2019.
9. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Data collection and calculation were performed independently by 2 blinded observers
(M.D. and J.O.) who had no therapeutic relationship with the patients.
All patients >15 years old with TBI who were treated at Ghent University Hospital
from January 2010 until December 2014 were included
Patients who were classified by the hospital emergency department under the term
“intracranial injury” were found eligible for analysis of the selection criteria.
Only patients with head trauma and neurologic deficits or with a skull fracture were
included
10. Department of Neurosurgery
Tribhuvan University Teaching Hospital
If this was not evident, patients who were admitted to the hospital for 24 hours or
patients who died within the first 24 hours were excluded.
Other exclusion criteria were age <15 years, no data available within the first 8 hours
after trauma, or a history of neurotrauma.
Mortality at 14 days after TBI was collected from the patient charts.
The GOS was used to categorize the outcome at 6 months (GOS score 1- death, GOS
score 5 - low disability).
11. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Prognostic factors of the CRASH calculator include
pupil reactivity,
GCS score,
CT scan findings (subdural hematoma, subarachnoid hematoma, petechial
bleeding, obliteration of the basal cisterns and third ventricle, midline shift, and
nonevacuated hematoma), and
extracranial injury.
13. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Extracranial injury defined as, that could lead to
hemodynamic instability,
Respiratory failure or
significant blood loss.
The CRASH application for TBI prognosis for Android v1.0 (Copyright 2013e2014,
Hyperexis, Sherbrooke, Quebec, Canada) calculated the chance of mortality at 14 days
after trauma and an unfavorable outcome at 6 months, which was defined as a GOS score
<4.
The chance of mortality was expressed as a percent score with the 95% confidence
interval (CI).
14. Department of Neurosurgery
Tribhuvan University Teaching Hospital
The patient cohort was split into 2 groups.
One observer studied one group concerning clinical outcome, and the CRASH score
was calculated by the other observer.
Data within the patient records could be found at different places using tabs.
Thus, the observer who calculated the CRASH score of a patient was blinded to the
outcome and vice versa.
The observers had their own data files to register their data.
After data collection, both data files were merged into the final database.
15. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Statistical analysis
Descriptive statistics were used for
patient characteristics,
cause of neurotrauma,
clinical features,
CT scan findings, and
information about neurosurgery
The calculator was validated using a receiver operating characteristic (ROC) curve.
The cutoff value with the highest combined sensitivity and specificity was determined.
16. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Next, this cutoff value was used to split the scores of the patients into 2 groups.
The positive predictive value, negative predictive value, and relative risk of both
cutoff scores were calculated using 2x2 table.
Logistic regression was used to calculate odds ratios.
Statistical analysis was performed using IBM SPSS Statistics Version 25 (IBM
Corporation, Armonk, New York, USA).
17. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Results
The study included 417 patients with TBI
The mean age of patients was 48 years (range, 15 to 88 years).
There were 292 male patients (70%) and 125 female patients (30%)
Cause
Fall- 50%
Motor vehicle accidents -33.1% of trauma.
Cycling 10%.
Intentional injuries- 10%
21. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Outcome statistics
Information at 14 days after trauma was collected in 396 patients (94.7%)
After this period, 334 patients (80.1%) were still alive.
The exact cause of death was reported in 25 of the 62 deaths (40.3%);
brain death was diagnosed in 20 patients (80%),
2 (12%) died of another neurologic cause, and
2 (8%) died of cardiac arrest.
22. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Of 56 patients who died at 14 days,
45 (40.5%) initially had a GCS score <8, and
11 (9.6%) had a GCS score between 8 and 14.
There were no patients with a GCS score of 15 who died after 14 days.
Information on the GOS score at 6 months was available for 307 patients (73.6%).
At 6 months, 61 patients (20%) were dead. Eleven patients (3.5%) were in a
persistent vegetative state.
23. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Among the remaining patients,
89 (29%) had a severe disability,
72 (23%) had a moderate disability, and
74 (24%) had a mild disability.
The mean CRASH scores of the patients who survived after 14 days and who died
were 12% and 59%, respectively.
The average CRASH score at 6 months of patients with a GOS score 4 was 28,
whereas the average score of patients with a GOS score <4 was 73%.
25. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Validation of CRASH calculator
The ROC curve showed that 14-day mortality risk (%) had a good prognostic value
The area under the curve (AUC) score obtained using the ROC method was 92.1%
(95% CI 88.5%-95.7%, P < 0.05).
The AUC for 14 days was 92.1%.
The cutoff value with the highest combined sensitivity and specificity was 31.5%
(0.823 sensitivity and 0.895 specificity).
26. Department of Neurosurgery
Tribhuvan University Teaching Hospital
The CRASH score was recoded into a dummy variable according to the cutoff value
of 31.5, and patients were split into 2 groups. The positive predictive value of this
cutoff value was 59.3%, and the negative predictive value was 96.5%.
If the calculated score was <31.5%, the chance for a patient to survive at 14 days was
96.5% (negative predictive value)
If the score was equal to or exceeded this score, the chance of mortality at 14 days
was 59.3% (positive predictive value).
Patients with a score of 31.5% were 16.7 times more likely to die at 2 weeks (relative
risk 16.7).
28. Department of Neurosurgery
Tribhuvan University Teaching Hospital
The same method was used to calculate the AUC for the CRASH score after 6
months (AUC 90.7%)
The cutoff value with the highest combined sensitivity and specificity was 55.75%
(0.793 sensitivity and 0.830 specificity).
The positive predictive value was 71.5%, and the negative predictive value was
88.1%.
This means that patients with a CRASH score 55.75% were 6 times more likely to
have a GOS score <4 after 6 months
31. Department of Neurosurgery
Tribhuvan University Teaching Hospital
The validity of the CRASH calculator was also tested by using logistic regression.
This showed that an increase of 1% on the CRASH calculator significantly correlated
with
an increase in chance of mortality after 14 days (odds ratio 1.076, 95% CI
1.059- 1.093) and
Increase in chance of a GOS score <4 at 6 months (odds ratio 1.070, 95% CI
1.055 -1.085).
32. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Discussion
The most frequent cause of TBI in this trial was falls, and the second most frequent
was traffic accidents.
Of the patients who had a traffic accident, 81.9% were male, and only 18.1% were
female, with a male-to-female ratio of 9:2
The fall occurs mostly in old ages, and hence the mean age is towards elderly.
No patient with a GCS score of 15 was dead after 14 days
33. Department of Neurosurgery
Tribhuvan University Teaching Hospital
To validate the CRASH clinical calculator, a cutoff value was chosen based on ROC
analysis
This means that if the CRASH calculator predicts that the patient will not survive the
first 14 days of neurotrauma, there is a 10.5% chance that this prediction is false.
Concerning the ROC curve of the score after 6 months, the value with the highest
specificity was chosen because the aim of this study was to keep the number of false-
positive results at a minimum.
The chosen cutoff value was 55.75%.
34. Department of Neurosurgery
Tribhuvan University Teaching Hospital
There is currently no consensus about whether the CRASH calculator is valid enough
to be used in daily clinical practice and which cutoff score should be used.
Heterogeneity is evident in the different studies concerning inclusion criteria, size of
study population, and conclusion on the validity of this calculator
In the present study, the inclusion criteria were as similar as possible to the inclusion
criteria used in the original CRASH trial.
A search of current literature about this topic yielded no other studies that included
patients with a GCS score of 15
35. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Han et al chose to externally validate only patients with severe TBI because this
group of patients could possibly benefit the most from an accurate prognostic model.
Several other studies also focused only on patients with severe TBI
As overestimation of an unfavorable outcome is an important finding throughout
multiple other studies, including a control group of patients with a GCS score of 15
could avoid an important bias
36. Department of Neurosurgery
Tribhuvan University Teaching Hospital
The study by Kwok et al. included 661 patients with TBI, of whom 310 patients met
the original CRASH inclusion criteria.
For this selected group of patients, the CRASH prognostic model made an accurate
estimation of the 14-day mortality and a mild underestimation of unfavorable
outcome at 6 months.
A recent review that included 58 studies published between 2006 and 2018
concluded that the CRASH calculator has adequate discriminative ability across a
range of settings (Dijkland SA et al.)
Implementation in clinical practice is recommended, provided that the calculator has
been validated or updated for the specific clinical setting.
37. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Critical appraisal
Strength
Good sample size
Inclusion criteria similar to original CRASH model
Good statistical analysis done using ROC and AUC
Inclusion of GCS 15 might have reduced overestimation of bad outcome and
mortality, and the study showed less poor outcome in this group
Observers were blinded hence observer bias reduced
38. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Critical appraisal
Weakness
Retrospective
Inclusion criteria included GCS 15, which might underestimate the risk
GOS 3 and 4 were difficult to interpret after 6 months, and was researcher
dependent
Does not take account in interpreting CT findings
Selected cutoff did not reach 100%, and hence chance of false positive and
negative results
39. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Conclusion
This retrospective study comprising 417 patients with TBI shows that the CRASH
prognostic calculator can be used in clinical practice if the patient matches the
original inclusion criteria of the calculator.
Owing to a false-negative predictive score of 3.5% at 14 days and 11.9% at 6 months
and no cutoff value with a combined sensitivity and specificity of 100%, the
calculator cannot replace the clinical decision-making process of physicians, but it
may serve as an ancillary tool.
Further research on clinical calculators is recommended.
40. Department of Neurosurgery
Tribhuvan University Teaching Hospital
Take home message
CRASH tool can be important tool for prediction of outcome in TBI as shown in
other clinical set up’s.
But it requires validation in our clinical setting to use it in daily practice.
It can be an important area of research in our set up.